smpboot.c 31.9 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
1
2
3
4
5
6
7
8
9
10
11
12
13
14
/*
 *	x86 SMP booting functions
 *
 *	(c) 1995 Alan Cox, Building #3 <alan@redhat.com>
 *	(c) 1998, 1999, 2000 Ingo Molnar <mingo@redhat.com>
 *	Copyright 2001 Andi Kleen, SuSE Labs.
 *
 *	Much of the core SMP work is based on previous work by Thomas Radke, to
 *	whom a great many thanks are extended.
 *
 *	Thanks to Intel for making available several different Pentium,
 *	Pentium Pro and Pentium-II/Xeon MP machines.
 *	Original development of Linux SMP code supported by Caldera.
 *
15
 *	This code is released under the GNU General Public License version 2
Linus Torvalds's avatar
Linus Torvalds committed
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
 *
 *	Fixes
 *		Felix Koop	:	NR_CPUS used properly
 *		Jose Renau	:	Handle single CPU case.
 *		Alan Cox	:	By repeated request 8) - Total BogoMIP report.
 *		Greg Wright	:	Fix for kernel stacks panic.
 *		Erich Boleyn	:	MP v1.4 and additional changes.
 *	Matthias Sattler	:	Changes for 2.1 kernel map.
 *	Michel Lespinasse	:	Changes for 2.1 kernel map.
 *	Michael Chastain	:	Change trampoline.S to gnu as.
 *		Alan Cox	:	Dumb bug: 'B' step PPro's are fine
 *		Ingo Molnar	:	Added APIC timers, based on code
 *					from Jose Renau
 *		Ingo Molnar	:	various cleanups and rewrites
 *		Tigran Aivazian	:	fixed "0.00 in /proc/uptime on SMP" bug.
 *	Maciej W. Rozycki	:	Bits for genuine 82489DX APICs
 *	Andi Kleen		:	Changed for SMP boot into long mode.
33
34
35
36
 *		Rusty Russell	:	Hacked into shape for new "hotplug" boot process.
 *      Andi Kleen              :       Converted to new state machine.
 *					Various cleanups.
 *					Probably mostly hotplug CPU ready now.
37
 *	Ashok Raj			: CPU hotplug support
Linus Torvalds's avatar
Linus Torvalds committed
38
39
 */

40

Linus Torvalds's avatar
Linus Torvalds committed
41
42
43
44
45
46
47
48
49
50
#include <linux/init.h>

#include <linux/mm.h>
#include <linux/kernel_stat.h>
#include <linux/smp_lock.h>
#include <linux/bootmem.h>
#include <linux/thread_info.h>
#include <linux/module.h>
#include <linux/delay.h>
#include <linux/mc146818rtc.h>
51
52
#include <linux/smp.h>

Linus Torvalds's avatar
Linus Torvalds committed
53
54
55
56
57
58
#include <asm/mtrr.h>
#include <asm/pgalloc.h>
#include <asm/desc.h>
#include <asm/kdebug.h>
#include <asm/tlbflush.h>
#include <asm/proto.h>
59
#include <asm/nmi.h>
60
61
#include <asm/irq.h>
#include <asm/hw_irq.h>
62
#include <asm/numa.h>
Linus Torvalds's avatar
Linus Torvalds committed
63
64
65

/* Number of siblings per CPU package */
int smp_num_siblings = 1;
66
EXPORT_SYMBOL(smp_num_siblings);
Linus Torvalds's avatar
Linus Torvalds committed
67

68
69
/* Last level cache ID of each logical CPU */
u8 cpu_llc_id[NR_CPUS] __cpuinitdata  = {[0 ... NR_CPUS-1] = BAD_APICID};
70
EXPORT_SYMBOL(cpu_llc_id);
71

Linus Torvalds's avatar
Linus Torvalds committed
72
/* Bitmask of currently online CPUs */
73
cpumask_t cpu_online_map __read_mostly;
Linus Torvalds's avatar
Linus Torvalds committed
74

75
76
77
78
79
80
EXPORT_SYMBOL(cpu_online_map);

/*
 * Private maps to synchronize booting between AP and BP.
 * Probably not needed anymore, but it makes for easier debugging. -AK
 */
Linus Torvalds's avatar
Linus Torvalds committed
81
82
cpumask_t cpu_callin_map;
cpumask_t cpu_callout_map;
83
EXPORT_SYMBOL(cpu_callout_map);
84
85
86

cpumask_t cpu_possible_map;
EXPORT_SYMBOL(cpu_possible_map);
Linus Torvalds's avatar
Linus Torvalds committed
87
88
89

/* Per CPU bogomips and other parameters */
struct cpuinfo_x86 cpu_data[NR_CPUS] __cacheline_aligned;
90
EXPORT_SYMBOL(cpu_data);
Linus Torvalds's avatar
Linus Torvalds committed
91

92
93
94
/* Set when the idlers are all forked */
int smp_threads_ready;

95
/* representing HT siblings of each logical CPU */
96
cpumask_t cpu_sibling_map[NR_CPUS] __read_mostly;
97
EXPORT_SYMBOL(cpu_sibling_map);
98
99

/* representing HT and core siblings of each logical CPU */
100
cpumask_t cpu_core_map[NR_CPUS] __read_mostly;
101
EXPORT_SYMBOL(cpu_core_map);
Linus Torvalds's avatar
Linus Torvalds committed
102
103
104
105
106

/*
 * Trampoline 80x86 program as an array.
 */

107
108
extern unsigned char trampoline_data[];
extern unsigned char trampoline_end[];
Linus Torvalds's avatar
Linus Torvalds committed
109

110
111
112
113
114
115
116
117
118
119
120
121
122
/* State of each CPU */
DEFINE_PER_CPU(int, cpu_state) = { 0 };

/*
 * Store all idle threads, this can be reused instead of creating
 * a new thread. Also avoids complicated thread destroy functionality
 * for idle threads.
 */
struct task_struct *idle_thread_array[NR_CPUS] __cpuinitdata ;

#define get_idle_for_cpu(x)     (idle_thread_array[(x)])
#define set_idle_for_cpu(x,p)   (idle_thread_array[(x)] = (p))

Linus Torvalds's avatar
Linus Torvalds committed
123
124
125
126
127
128
/*
 * Currently trivial. Write the real->protected mode
 * bootstrap into the page concerned. The caller
 * has made sure it's suitably aligned.
 */

129
static unsigned long __cpuinit setup_trampoline(void)
Linus Torvalds's avatar
Linus Torvalds committed
130
131
132
133
134
135
136
137
138
139
140
{
	void *tramp = __va(SMP_TRAMPOLINE_BASE); 
	memcpy(tramp, trampoline_data, trampoline_end - trampoline_data);
	return virt_to_phys(tramp);
}

/*
 * The bootstrap kernel entry code has set these up. Save them for
 * a given CPU
 */

141
static void __cpuinit smp_store_cpu_info(int id)
Linus Torvalds's avatar
Linus Torvalds committed
142
143
144
145
146
{
	struct cpuinfo_x86 *c = cpu_data + id;

	*c = boot_cpu_data;
	identify_cpu(c);
147
	print_cpu_info(c);
Linus Torvalds's avatar
Linus Torvalds committed
148
149
150
}

/*
151
152
153
154
155
 * New Funky TSC sync algorithm borrowed from IA64.
 * Main advantage is that it doesn't reset the TSCs fully and
 * in general looks more robust and it works better than my earlier
 * attempts. I believe it was written by David Mosberger. Some minor
 * adjustments for x86-64 by me -AK
Linus Torvalds's avatar
Linus Torvalds committed
156
 *
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
 * Original comment reproduced below.
 *
 * Synchronize TSC of the current (slave) CPU with the TSC of the
 * MASTER CPU (normally the time-keeper CPU).  We use a closed loop to
 * eliminate the possibility of unaccounted-for errors (such as
 * getting a machine check in the middle of a calibration step).  The
 * basic idea is for the slave to ask the master what itc value it has
 * and to read its own itc before and after the master responds.  Each
 * iteration gives us three timestamps:
 *
 *	slave		master
 *
 *	t0 ---\
 *             ---\
 *		   --->
 *			tm
 *		   /---
 *	       /---
 *	t1 <---
 *
 *
 * The goal is to adjust the slave's TSC such that tm falls exactly
 * half-way between t0 and t1.  If we achieve this, the clocks are
 * synchronized provided the interconnect between the slave and the
 * master is symmetric.  Even if the interconnect were asymmetric, we
 * would still know that the synchronization error is smaller than the
 * roundtrip latency (t0 - t1).
 *
 * When the interconnect is quiet and symmetric, this lets us
 * synchronize the TSC to within one or two cycles.  However, we can
 * only *guarantee* that the synchronization is accurate to within a
 * round-trip time, which is typically in the range of several hundred
 * cycles (e.g., ~500 cycles).  In practice, this means that the TSCs
 * are usually almost perfectly synchronized, but we shouldn't assume
 * that the accuracy is much better than half a micro second or so.
 *
 * [there are other errors like the latency of RDTSC and of the
 * WRMSR. These can also account to hundreds of cycles. So it's
 * probably worse. It claims 153 cycles error on a dual Opteron,
 * but I suspect the numbers are actually somewhat worse -AK]
Linus Torvalds's avatar
Linus Torvalds committed
197
198
 */

199
200
201
202
203
204
205
206
207
#define MASTER	0
#define SLAVE	(SMP_CACHE_BYTES/8)

/* Intentionally don't use cpu_relax() while TSC synchronization
   because we don't want to go into funky power save modi or cause
   hypervisors to schedule us away.  Going to sleep would likely affect
   latency and low latency is the primary objective here. -AK */
#define no_cpu_relax() barrier()

208
static __cpuinitdata DEFINE_SPINLOCK(tsc_sync_lock);
209
210
211
212
static volatile __cpuinitdata unsigned long go[SLAVE + 1];
static int notscsync __cpuinitdata;

#undef DEBUG_TSC_SYNC
Linus Torvalds's avatar
Linus Torvalds committed
213

214
215
#define NUM_ROUNDS	64	/* magic value */
#define NUM_ITERS	5	/* likewise */
Linus Torvalds's avatar
Linus Torvalds committed
216

217
218
/* Callback on boot CPU */
static __cpuinit void sync_master(void *arg)
Linus Torvalds's avatar
Linus Torvalds committed
219
{
220
221
222
223
224
225
226
227
228
229
230
231
232
233
	unsigned long flags, i;

	go[MASTER] = 0;

	local_irq_save(flags);
	{
		for (i = 0; i < NUM_ROUNDS*NUM_ITERS; ++i) {
			while (!go[MASTER])
				no_cpu_relax();
			go[MASTER] = 0;
			rdtscll(go[SLAVE]);
		}
	}
	local_irq_restore(flags);
234
}
Linus Torvalds's avatar
Linus Torvalds committed
235

236
/*
237
238
239
 * Return the number of cycles by which our tsc differs from the tsc
 * on the master (time-keeper) CPU.  A positive number indicates our
 * tsc is ahead of the master, negative that it is behind.
240
 */
241
242
static inline long
get_delta(long *rt, long *master)
243
{
244
245
246
	unsigned long best_t0 = 0, best_t1 = ~0UL, best_tm = 0;
	unsigned long tcenter, t0, t1, tm;
	int i;
247

248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
	for (i = 0; i < NUM_ITERS; ++i) {
		rdtscll(t0);
		go[MASTER] = 1;
		while (!(tm = go[SLAVE]))
			no_cpu_relax();
		go[SLAVE] = 0;
		rdtscll(t1);

		if (t1 - t0 < best_t1 - best_t0)
			best_t0 = t0, best_t1 = t1, best_tm = tm;
	}

	*rt = best_t1 - best_t0;
	*master = best_tm - best_t0;

	/* average best_t0 and best_t1 without overflow: */
	tcenter = (best_t0/2 + best_t1/2);
	if (best_t0 % 2 + best_t1 % 2 == 2)
		++tcenter;
	return tcenter - best_tm;
Linus Torvalds's avatar
Linus Torvalds committed
268
269
}

Eric W. Biederman's avatar
Eric W. Biederman committed
270
static __cpuinit void sync_tsc(unsigned int master)
Linus Torvalds's avatar
Linus Torvalds committed
271
{
272
273
274
	int i, done = 0;
	long delta, adj, adjust_latency = 0;
	unsigned long flags, rt, master_time_stamp, bound;
275
#ifdef DEBUG_TSC_SYNC
276
277
278
279
280
281
282
283
	static struct syncdebug {
		long rt;	/* roundtrip time */
		long master;	/* master's timestamp */
		long diff;	/* difference between midpoint and master's timestamp */
		long lat;	/* estimate of tsc adjustment latency */
	} t[NUM_ROUNDS] __cpuinitdata;
#endif

Eric W. Biederman's avatar
Eric W. Biederman committed
284
285
286
	printk(KERN_INFO "CPU %d: Syncing TSC to CPU %u.\n",
		smp_processor_id(), master);

287
288
	go[MASTER] = 1;

Eric W. Biederman's avatar
Eric W. Biederman committed
289
290
291
292
293
294
	/* It is dangerous to broadcast IPI as cpus are coming up,
	 * as they may not be ready to accept them.  So since
	 * we only need to send the ipi to the boot cpu direct
	 * the message, and avoid the race.
	 */
	smp_call_function_single(master, sync_master, NULL, 1, 0);
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318

	while (go[MASTER])	/* wait for master to be ready */
		no_cpu_relax();

	spin_lock_irqsave(&tsc_sync_lock, flags);
	{
		for (i = 0; i < NUM_ROUNDS; ++i) {
			delta = get_delta(&rt, &master_time_stamp);
			if (delta == 0) {
				done = 1;	/* let's lock on to this... */
				bound = rt;
			}

			if (!done) {
				unsigned long t;
				if (i > 0) {
					adjust_latency += -delta;
					adj = -delta + adjust_latency/4;
				} else
					adj = -delta;

				rdtscll(t);
				wrmsrl(MSR_IA32_TSC, t + adj);
			}
319
#ifdef DEBUG_TSC_SYNC
320
321
322
323
324
325
326
327
328
			t[i].rt = rt;
			t[i].master = master_time_stamp;
			t[i].diff = delta;
			t[i].lat = adjust_latency/4;
#endif
		}
	}
	spin_unlock_irqrestore(&tsc_sync_lock, flags);

329
#ifdef DEBUG_TSC_SYNC
330
331
332
333
334
335
336
337
	for (i = 0; i < NUM_ROUNDS; ++i)
		printk("rt=%5ld master=%5ld diff=%5ld adjlat=%5ld\n",
		       t[i].rt, t[i].master, t[i].diff, t[i].lat);
#endif

	printk(KERN_INFO
	       "CPU %d: synchronized TSC with CPU %u (last diff %ld cycles, "
	       "maxerr %lu cycles)\n",
Eric W. Biederman's avatar
Eric W. Biederman committed
338
	       smp_processor_id(), master, delta, rt);
339
}
Linus Torvalds's avatar
Linus Torvalds committed
340

341
static void __cpuinit tsc_sync_wait(void)
342
{
343
344
345
346
347
348
349
	/*
	 * When the CPU has synchronized TSCs assume the BIOS
  	 * or the hardware already synced.  Otherwise we could
	 * mess up a possible perfect synchronization with a
	 * not-quite-perfect algorithm.
	 */
	if (notscsync || !cpu_has_tsc || !unsynchronized_tsc())
350
		return;
351
	sync_tsc(0);
352
}
Linus Torvalds's avatar
Linus Torvalds committed
353

354
static __init int notscsync_setup(char *s)
355
{
356
	notscsync = 1;
357
	return 1;
Linus Torvalds's avatar
Linus Torvalds committed
358
}
359
__setup("notscsync", notscsync_setup);
Linus Torvalds's avatar
Linus Torvalds committed
360

361
static atomic_t init_deasserted __cpuinitdata;
Linus Torvalds's avatar
Linus Torvalds committed
362

363
364
365
366
367
/*
 * Report back to the Boot Processor.
 * Running on AP.
 */
void __cpuinit smp_callin(void)
Linus Torvalds's avatar
Linus Torvalds committed
368
369
370
371
372
373
374
375
376
377
{
	int cpuid, phys_id;
	unsigned long timeout;

	/*
	 * If waken up by an INIT in an 82489DX configuration
	 * we may get here before an INIT-deassert IPI reaches
	 * our local APIC.  We have to wait for the IPI or we'll
	 * lock up on an APIC access.
	 */
378
379
	while (!atomic_read(&init_deasserted))
		cpu_relax();
Linus Torvalds's avatar
Linus Torvalds committed
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409

	/*
	 * (This works even if the APIC is not enabled.)
	 */
	phys_id = GET_APIC_ID(apic_read(APIC_ID));
	cpuid = smp_processor_id();
	if (cpu_isset(cpuid, cpu_callin_map)) {
		panic("smp_callin: phys CPU#%d, CPU#%d already present??\n",
					phys_id, cpuid);
	}
	Dprintk("CPU#%d (phys ID: %d) waiting for CALLOUT\n", cpuid, phys_id);

	/*
	 * STARTUP IPIs are fragile beasts as they might sometimes
	 * trigger some glue motherboard logic. Complete APIC bus
	 * silence for 1 second, this overestimates the time the
	 * boot CPU is spending to send the up to 2 STARTUP IPIs
	 * by a factor of two. This should be enough.
	 */

	/*
	 * Waiting 2s total for startup (udelay is not yet working)
	 */
	timeout = jiffies + 2*HZ;
	while (time_before(jiffies, timeout)) {
		/*
		 * Has the boot CPU finished it's STARTUP sequence?
		 */
		if (cpu_isset(cpuid, cpu_callout_map))
			break;
410
		cpu_relax();
Linus Torvalds's avatar
Linus Torvalds committed
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
	}

	if (!time_before(jiffies, timeout)) {
		panic("smp_callin: CPU%d started up but did not get a callout!\n",
			cpuid);
	}

	/*
	 * the boot CPU has finished the init stage and is spinning
	 * on callin_map until we finish. We are free to set up this
	 * CPU, first the APIC. (this is probably redundant on most
	 * boards)
	 */

	Dprintk("CALLIN, before setup_local_APIC().\n");
	setup_local_APIC();

	/*
	 * Get our bogomips.
430
431
432
 	 *
 	 * Need to enable IRQs because it can take longer and then
	 * the NMI watchdog might kill us.
Linus Torvalds's avatar
Linus Torvalds committed
433
	 */
434
	local_irq_enable();
Linus Torvalds's avatar
Linus Torvalds committed
435
	calibrate_delay();
436
	local_irq_disable();
Linus Torvalds's avatar
Linus Torvalds committed
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
	Dprintk("Stack at about %p\n",&cpuid);

	disable_APIC_timer();

	/*
	 * Save our processor parameters
	 */
 	smp_store_cpu_info(cpuid);

	/*
	 * Allow the master to continue.
	 */
	cpu_set(cpuid, cpu_callin_map);
}

452
453
454
455
456
457
/* maps the cpu to the sched domain representing multi-core */
cpumask_t cpu_coregroup_map(int cpu)
{
	struct cpuinfo_x86 *c = cpu_data + cpu;
	/*
	 * For perf, we return last level cache shared map.
458
	 * And for power savings, we return cpu_core_map
459
	 */
460
461
462
463
	if (sched_mc_power_savings || sched_smt_power_savings)
		return cpu_core_map[cpu];
	else
		return c->llc_shared_map;
464
465
}

466
467
468
/* representing cpus for which sibling maps can be computed */
static cpumask_t cpu_sibling_setup_map;

469
470
471
static inline void set_cpu_sibling_map(int cpu)
{
	int i;
472
473
474
	struct cpuinfo_x86 *c = cpu_data;

	cpu_set(cpu, cpu_sibling_setup_map);
475
476

	if (smp_num_siblings > 1) {
477
		for_each_cpu_mask(i, cpu_sibling_setup_map) {
478
479
			if (c[cpu].phys_proc_id == c[i].phys_proc_id &&
			    c[cpu].cpu_core_id == c[i].cpu_core_id) {
480
481
				cpu_set(i, cpu_sibling_map[cpu]);
				cpu_set(cpu, cpu_sibling_map[i]);
482
483
				cpu_set(i, cpu_core_map[cpu]);
				cpu_set(cpu, cpu_core_map[i]);
484
485
				cpu_set(i, c[cpu].llc_shared_map);
				cpu_set(cpu, c[i].llc_shared_map);
486
487
488
489
490
491
			}
		}
	} else {
		cpu_set(cpu, cpu_sibling_map[cpu]);
	}

492
493
	cpu_set(cpu, c[cpu].llc_shared_map);

494
	if (current_cpu_data.x86_max_cores == 1) {
495
		cpu_core_map[cpu] = cpu_sibling_map[cpu];
496
497
498
499
500
		c[cpu].booted_cores = 1;
		return;
	}

	for_each_cpu_mask(i, cpu_sibling_setup_map) {
501
502
503
504
505
		if (cpu_llc_id[cpu] != BAD_APICID &&
		    cpu_llc_id[cpu] == cpu_llc_id[i]) {
			cpu_set(i, c[cpu].llc_shared_map);
			cpu_set(cpu, c[i].llc_shared_map);
		}
506
		if (c[cpu].phys_proc_id == c[i].phys_proc_id) {
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
			cpu_set(i, cpu_core_map[cpu]);
			cpu_set(cpu, cpu_core_map[i]);
			/*
			 *  Does this new cpu bringup a new core?
			 */
			if (cpus_weight(cpu_sibling_map[cpu]) == 1) {
				/*
				 * for each core in package, increment
				 * the booted_cores for this new cpu
				 */
				if (first_cpu(cpu_sibling_map[i]) == i)
					c[cpu].booted_cores++;
				/*
				 * increment the core count for all
				 * the other cpus in this package
				 */
				if (i != cpu)
					c[i].booted_cores++;
			} else if (i != cpu && !c[cpu].booted_cores)
				c[cpu].booted_cores = c[i].booted_cores;
		}
528
529
530
	}
}

Linus Torvalds's avatar
Linus Torvalds committed
531
/*
532
 * Setup code on secondary processor (after comming out of the trampoline)
Linus Torvalds's avatar
Linus Torvalds committed
533
 */
534
void __cpuinit start_secondary(void)
Linus Torvalds's avatar
Linus Torvalds committed
535
536
537
538
539
540
541
{
	/*
	 * Dont put anything before smp_callin(), SMP
	 * booting is too fragile that we want to limit the
	 * things done here to the most necessary things.
	 */
	cpu_init();
542
	preempt_disable();
Linus Torvalds's avatar
Linus Torvalds committed
543
544
545
546
547
548
549
550
	smp_callin();

	/* otherwise gcc will move up the smp_processor_id before the cpu_init */
	barrier();

	Dprintk("cpu %d: setting up apic clock\n", smp_processor_id()); 	
	setup_secondary_APIC_clock();

551
	Dprintk("cpu %d: enabling apic timer\n", smp_processor_id());
Linus Torvalds's avatar
Linus Torvalds committed
552
553
554
555
556
557
558

	if (nmi_watchdog == NMI_IO_APIC) {
		disable_8259A_irq(0);
		enable_NMI_through_LVT0(NULL);
		enable_8259A_irq(0);
	}

559
	enable_APIC_timer();
Linus Torvalds's avatar
Linus Torvalds committed
560

561
562
563
564
565
566
	/*
	 * The sibling maps must be set before turing the online map on for
	 * this cpu
	 */
	set_cpu_sibling_map(smp_processor_id());

567
568
569
570
571
572
573
574
	/* 
  	 * Wait for TSC sync to not schedule things before.
	 * We still process interrupts, which could see an inconsistent
	 * time in that window unfortunately. 
	 * Do this here because TSC sync has global unprotected state.
 	 */
	tsc_sync_wait();

575
576
577
578
579
580
581
582
583
	/*
	 * We need to hold call_lock, so there is no inconsistency
	 * between the time smp_call_function() determines number of
	 * IPI receipients, and the time when the determination is made
	 * for which cpus receive the IPI in genapic_flat.c. Holding this
	 * lock helps us to not include this cpu in a currently in progress
	 * smp_call_function().
	 */
	lock_ipi_call_lock();
584
	spin_lock(&vector_lock);
585

586
587
	/* Setup the per cpu irq handling data structures */
	__setup_vector_irq(smp_processor_id());
Linus Torvalds's avatar
Linus Torvalds committed
588
	/*
589
	 * Allow the master to continue.
Linus Torvalds's avatar
Linus Torvalds committed
590
591
	 */
	cpu_set(smp_processor_id(), cpu_online_map);
592
	per_cpu(cpu_state, smp_processor_id()) = CPU_ONLINE;
593
	spin_unlock(&vector_lock);
594
595
	unlock_ipi_call_lock();

Linus Torvalds's avatar
Linus Torvalds committed
596
597
598
	cpu_idle();
}

599
extern volatile unsigned long init_rsp;
Linus Torvalds's avatar
Linus Torvalds committed
600
601
extern void (*initial_code)(void);

602
#ifdef APIC_DEBUG
603
static void inquire_remote_apic(int apicid)
Linus Torvalds's avatar
Linus Torvalds committed
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
{
	unsigned i, regs[] = { APIC_ID >> 4, APIC_LVR >> 4, APIC_SPIV >> 4 };
	char *names[] = { "ID", "VERSION", "SPIV" };
	int timeout, status;

	printk(KERN_INFO "Inquiring remote APIC #%d...\n", apicid);

	for (i = 0; i < sizeof(regs) / sizeof(*regs); i++) {
		printk("... APIC #%d %s: ", apicid, names[i]);

		/*
		 * Wait for idle.
		 */
		apic_wait_icr_idle();

619
620
		apic_write(APIC_ICR2, SET_APIC_DEST_FIELD(apicid));
		apic_write(APIC_ICR, APIC_DM_REMRD | regs[i]);
Linus Torvalds's avatar
Linus Torvalds committed
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639

		timeout = 0;
		do {
			udelay(100);
			status = apic_read(APIC_ICR) & APIC_ICR_RR_MASK;
		} while (status == APIC_ICR_RR_INPROG && timeout++ < 1000);

		switch (status) {
		case APIC_ICR_RR_VALID:
			status = apic_read(APIC_RRR);
			printk("%08x\n", status);
			break;
		default:
			printk("failed\n");
		}
	}
}
#endif

640
641
642
643
/*
 * Kick the secondary to wake up.
 */
static int __cpuinit wakeup_secondary_via_INIT(int phys_apicid, unsigned int start_rip)
Linus Torvalds's avatar
Linus Torvalds committed
644
645
646
647
648
649
650
651
652
{
	unsigned long send_status = 0, accept_status = 0;
	int maxlvt, timeout, num_starts, j;

	Dprintk("Asserting INIT.\n");

	/*
	 * Turn INIT on target chip
	 */
653
	apic_write(APIC_ICR2, SET_APIC_DEST_FIELD(phys_apicid));
Linus Torvalds's avatar
Linus Torvalds committed
654
655
656
657

	/*
	 * Send IPI
	 */
658
	apic_write(APIC_ICR, APIC_INT_LEVELTRIG | APIC_INT_ASSERT
Linus Torvalds's avatar
Linus Torvalds committed
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
				| APIC_DM_INIT);

	Dprintk("Waiting for send to finish...\n");
	timeout = 0;
	do {
		Dprintk("+");
		udelay(100);
		send_status = apic_read(APIC_ICR) & APIC_ICR_BUSY;
	} while (send_status && (timeout++ < 1000));

	mdelay(10);

	Dprintk("Deasserting INIT.\n");

	/* Target chip */
674
	apic_write(APIC_ICR2, SET_APIC_DEST_FIELD(phys_apicid));
Linus Torvalds's avatar
Linus Torvalds committed
675
676

	/* Send IPI */
677
	apic_write(APIC_ICR, APIC_INT_LEVELTRIG | APIC_DM_INIT);
Linus Torvalds's avatar
Linus Torvalds committed
678
679
680
681
682
683
684
685
686

	Dprintk("Waiting for send to finish...\n");
	timeout = 0;
	do {
		Dprintk("+");
		udelay(100);
		send_status = apic_read(APIC_ICR) & APIC_ICR_BUSY;
	} while (send_status && (timeout++ < 1000));

687
	mb();
Linus Torvalds's avatar
Linus Torvalds committed
688
689
	atomic_set(&init_deasserted, 1);

690
	num_starts = 2;
Linus Torvalds's avatar
Linus Torvalds committed
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709

	/*
	 * Run STARTUP IPI loop.
	 */
	Dprintk("#startup loops: %d.\n", num_starts);

	maxlvt = get_maxlvt();

	for (j = 1; j <= num_starts; j++) {
		Dprintk("Sending STARTUP #%d.\n",j);
		apic_write(APIC_ESR, 0);
		apic_read(APIC_ESR);
		Dprintk("After apic_write.\n");

		/*
		 * STARTUP IPI
		 */

		/* Target chip */
710
		apic_write(APIC_ICR2, SET_APIC_DEST_FIELD(phys_apicid));
Linus Torvalds's avatar
Linus Torvalds committed
711
712
713

		/* Boot on the stack */
		/* Kick the second */
714
		apic_write(APIC_ICR, APIC_DM_STARTUP | (start_rip >> 12));
Linus Torvalds's avatar
Linus Torvalds committed
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754

		/*
		 * Give the other CPU some time to accept the IPI.
		 */
		udelay(300);

		Dprintk("Startup point 1.\n");

		Dprintk("Waiting for send to finish...\n");
		timeout = 0;
		do {
			Dprintk("+");
			udelay(100);
			send_status = apic_read(APIC_ICR) & APIC_ICR_BUSY;
		} while (send_status && (timeout++ < 1000));

		/*
		 * Give the other CPU some time to accept the IPI.
		 */
		udelay(200);
		/*
		 * Due to the Pentium erratum 3AP.
		 */
		if (maxlvt > 3) {
			apic_write(APIC_ESR, 0);
		}
		accept_status = (apic_read(APIC_ESR) & 0xEF);
		if (send_status || accept_status)
			break;
	}
	Dprintk("After Startup.\n");

	if (send_status)
		printk(KERN_ERR "APIC never delivered???\n");
	if (accept_status)
		printk(KERN_ERR "APIC delivery error (%lx).\n", accept_status);

	return (send_status | accept_status);
}

755
756
757
758
759
760
761
762
763
764
765
766
767
768
struct create_idle {
	struct task_struct *idle;
	struct completion done;
	int cpu;
};

void do_fork_idle(void *_c_idle)
{
	struct create_idle *c_idle = _c_idle;

	c_idle->idle = fork_idle(c_idle->cpu);
	complete(&c_idle->done);
}

769
770
771
772
/*
 * Boot one CPU.
 */
static int __cpuinit do_boot_cpu(int cpu, int apicid)
Linus Torvalds's avatar
Linus Torvalds committed
773
774
{
	unsigned long boot_error;
775
	int timeout;
Linus Torvalds's avatar
Linus Torvalds committed
776
	unsigned long start_rip;
777
778
	struct create_idle c_idle = {
		.cpu = cpu,
779
		.done = COMPLETION_INITIALIZER_ONSTACK(c_idle.done),
780
781
782
	};
	DECLARE_WORK(work, do_fork_idle, &c_idle);

783
784
785
786
787
788
789
	/* allocate memory for gdts of secondary cpus. Hotplug is considered */
	if (!cpu_gdt_descr[cpu].address &&
		!(cpu_gdt_descr[cpu].address = get_zeroed_page(GFP_KERNEL))) {
		printk(KERN_ERR "Failed to allocate GDT for CPU %d\n", cpu);
		return -1;
	}

790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
	/* Allocate node local memory for AP pdas */
	if (cpu_pda(cpu) == &boot_cpu_pda[cpu]) {
		struct x8664_pda *newpda, *pda;
		int node = cpu_to_node(cpu);
		pda = cpu_pda(cpu);
		newpda = kmalloc_node(sizeof (struct x8664_pda), GFP_ATOMIC,
				      node);
		if (newpda) {
			memcpy(newpda, pda, sizeof (struct x8664_pda));
			cpu_pda(cpu) = newpda;
		} else
			printk(KERN_ERR
		"Could not allocate node local PDA for CPU %d on node %d\n",
				cpu, node);
	}

806
807
	alternatives_smp_switch(1);

808
809
810
811
	c_idle.idle = get_idle_for_cpu(cpu);

	if (c_idle.idle) {
		c_idle.idle->thread.rsp = (unsigned long) (((struct pt_regs *)
Al Viro's avatar
Al Viro committed
812
			(THREAD_SIZE +  task_stack_page(c_idle.idle))) - 1);
813
814
815
816
		init_idle(c_idle.idle, cpu);
		goto do_rest;
	}

Linus Torvalds's avatar
Linus Torvalds committed
817
	/*
818
819
820
821
822
823
824
825
	 * During cold boot process, keventd thread is not spun up yet.
	 * When we do cpu hot-add, we create idle threads on the fly, we should
	 * not acquire any attributes from the calling context. Hence the clean
	 * way to create kernel_threads() is to do that from keventd().
	 * We do the current_is_keventd() due to the fact that ACPI notifier
	 * was also queuing to keventd() and when the caller is already running
	 * in context of keventd(), we would end up with locking up the keventd
	 * thread.
Linus Torvalds's avatar
Linus Torvalds committed
826
	 */
827
828
829
830
831
832
833
834
	if (!keventd_up() || current_is_keventd())
		work.func(work.data);
	else {
		schedule_work(&work);
		wait_for_completion(&c_idle.done);
	}

	if (IS_ERR(c_idle.idle)) {
835
		printk("failed fork for CPU %d\n", cpu);
836
		return PTR_ERR(c_idle.idle);
837
	}
Linus Torvalds's avatar
Linus Torvalds committed
838

839
840
841
842
	set_idle_for_cpu(cpu, c_idle.idle);

do_rest:

843
	cpu_pda(cpu)->pcurrent = c_idle.idle;
Linus Torvalds's avatar
Linus Torvalds committed
844
845
846

	start_rip = setup_trampoline();

847
	init_rsp = c_idle.idle->thread.rsp;
Linus Torvalds's avatar
Linus Torvalds committed
848
849
	per_cpu(init_tss,cpu).rsp0 = init_rsp;
	initial_code = start_secondary;
Al Viro's avatar
Al Viro committed
850
	clear_tsk_thread_flag(c_idle.idle, TIF_FORK);
Linus Torvalds's avatar
Linus Torvalds committed
851

852
853
854
	printk(KERN_INFO "Booting processor %d/%d APIC 0x%x\n", cpu,
		cpus_weight(cpu_present_map),
		apicid);
Linus Torvalds's avatar
Linus Torvalds committed
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875

	/*
	 * This grunge runs the startup process for
	 * the targeted processor.
	 */

	atomic_set(&init_deasserted, 0);

	Dprintk("Setting warm reset code and vector.\n");

	CMOS_WRITE(0xa, 0xf);
	local_flush_tlb();
	Dprintk("1.\n");
	*((volatile unsigned short *) phys_to_virt(0x469)) = start_rip >> 4;
	Dprintk("2.\n");
	*((volatile unsigned short *) phys_to_virt(0x467)) = start_rip & 0xf;
	Dprintk("3.\n");

	/*
	 * Be paranoid about clearing APIC errors.
	 */
876
877
	apic_write(APIC_ESR, 0);
	apic_read(APIC_ESR);
Linus Torvalds's avatar
Linus Torvalds committed
878
879
880
881
882
883
884
885
886

	/*
	 * Status is now clean
	 */
	boot_error = 0;

	/*
	 * Starting actual IPI sequence...
	 */
887
	boot_error = wakeup_secondary_via_INIT(apicid, start_rip);
Linus Torvalds's avatar
Linus Torvalds committed
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917

	if (!boot_error) {
		/*
		 * allow APs to start initializing.
		 */
		Dprintk("Before Callout %d.\n", cpu);
		cpu_set(cpu, cpu_callout_map);
		Dprintk("After Callout %d.\n", cpu);

		/*
		 * Wait 5s total for a response
		 */
		for (timeout = 0; timeout < 50000; timeout++) {
			if (cpu_isset(cpu, cpu_callin_map))
				break;	/* It has booted */
			udelay(100);
		}

		if (cpu_isset(cpu, cpu_callin_map)) {
			/* number CPUs logically, starting from 1 (BSP is 0) */
			Dprintk("CPU has booted.\n");
		} else {
			boot_error = 1;
			if (*((volatile unsigned char *)phys_to_virt(SMP_TRAMPOLINE_BASE))
					== 0xA5)
				/* trampoline started but...? */
				printk("Stuck ??\n");
			else
				/* trampoline code not run */
				printk("Not responding.\n");
918
#ifdef APIC_DEBUG
Linus Torvalds's avatar
Linus Torvalds committed
919
920
921
922
923
924
925
			inquire_remote_apic(apicid);
#endif
		}
	}
	if (boot_error) {
		cpu_clear(cpu, cpu_callout_map); /* was set here (do_boot_cpu()) */
		clear_bit(cpu, &cpu_initialized); /* was set by cpu_init() */
926
		clear_node_cpumask(cpu); /* was set by numa_add_cpu */
927
928
		cpu_clear(cpu, cpu_present_map);
		cpu_clear(cpu, cpu_possible_map);
Linus Torvalds's avatar
Linus Torvalds committed
929
930
		x86_cpu_to_apicid[cpu] = BAD_APICID;
		x86_cpu_to_log_apicid[cpu] = BAD_APICID;
931
		return -EIO;
Linus Torvalds's avatar
Linus Torvalds committed
932
	}
933
934

	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
935
936
}

937
938
939
cycles_t cacheflush_time;
unsigned long cache_decay_ticks;

Linus Torvalds's avatar
Linus Torvalds committed
940
/*
941
 * Cleanup possible dangling ends...
Linus Torvalds's avatar
Linus Torvalds committed
942
 */
943
static __cpuinit void smp_cleanup_boot(void)
Linus Torvalds's avatar
Linus Torvalds committed
944
{
945
946
947
948
949
	/*
	 * Paranoid:  Set warm reset code and vector here back
	 * to default values.
	 */
	CMOS_WRITE(0, 0xf);
Linus Torvalds's avatar
Linus Torvalds committed
950

951
952
953
954
955
956
957
958
959
960
961
	/*
	 * Reset trampoline flag
	 */
	*((volatile int *) phys_to_virt(0x467)) = 0;
}

/*
 * Fall back to non SMP mode after errors.
 *
 * RED-PEN audit/test this more. I bet there is more state messed up here.
 */
962
static __init void disable_smp(void)
963
964
965
966
967
968
969
970
971
972
973
{
	cpu_present_map = cpumask_of_cpu(0);
	cpu_possible_map = cpumask_of_cpu(0);
	if (smp_found_config)
		phys_cpu_present_map = physid_mask_of_physid(boot_cpu_id);
	else
		phys_cpu_present_map = physid_mask_of_physid(0);
	cpu_set(0, cpu_sibling_map[0]);
	cpu_set(0, cpu_core_map[0]);
}

974
#ifdef CONFIG_HOTPLUG_CPU
975
976
977

int additional_cpus __initdata = -1;

978
979
980
981
982
983
984
985
986
/*
 * cpu_possible_map should be static, it cannot change as cpu's
 * are onlined, or offlined. The reason is per-cpu data-structures
 * are allocated by some modules at init time, and dont expect to
 * do this dynamically on cpu arrival/departure.
 * cpu_present_map on the other hand can change dynamically.
 * In case when cpu_hotplug is not compiled, then we resort to current
 * behaviour, which is cpu_possible == cpu_present.
 * - Ashok Raj
987
988
989
990
 *
 * Three ways to find out the number of additional hotplug CPUs:
 * - If the BIOS specified disabled CPUs in ACPI/mptables use that.
 * - The user can overwrite it with additional_cpus=NUM
991
 * - Otherwise don't reserve additional CPUs.
992
993
 * We do this because additional CPUs waste a lot of memory.
 * -AK
994
 */
995
__init void prefill_possible_map(void)
996
997
{
	int i;
998
999
1000
	int possible;

 	if (additional_cpus == -1) {
1001
 		if (disabled_cpus > 0)
1002
 			additional_cpus = disabled_cpus;
1003
1004
 		else
			additional_cpus = 0;
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
 	}
	possible = num_processors + additional_cpus;
	if (possible > NR_CPUS) 
		possible = NR_CPUS;

	printk(KERN_INFO "SMP: Allowing %d CPUs, %d hotplug CPUs\n",
		possible,
	        max_t(int, possible - num_processors, 0));

	for (i = 0; i < possible; i++)
1015
1016
1017
1018
		cpu_set(i, cpu_possible_map);
}
#endif

1019
1020
1021
/*
 * Various sanity checks.
 */
1022
static int __init smp_sanity_check(unsigned max_cpus)
1023
{
Linus Torvalds's avatar
Linus Torvalds committed
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
	if (!physid_isset(hard_smp_processor_id(), phys_cpu_present_map)) {
		printk("weird, boot CPU (#%d) not listed by the BIOS.\n",
		       hard_smp_processor_id());
		physid_set(hard_smp_processor_id(), phys_cpu_present_map);
	}

	/*
	 * If we couldn't find an SMP configuration at boot time,
	 * get out of here now!
	 */
	if (!smp_found_config) {
		printk(KERN_NOTICE "SMP motherboard not detected.\n");
1036
		disable_smp();
Linus Torvalds's avatar
Linus Torvalds committed
1037
1038
1039
		if (APIC_init_uniprocessor())
			printk(KERN_NOTICE "Local APIC not detected."
					   " Using dummy APIC emulation.\n");
1040
		return -1;
Linus Torvalds's avatar
Linus Torvalds committed
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
	}

	/*
	 * Should not be necessary because the MP table should list the boot
	 * CPU too, but we do it for the sake of robustness anyway.
	 */
	if (!physid_isset(boot_cpu_id, phys_cpu_present_map)) {
		printk(KERN_NOTICE "weird, boot CPU (#%d) not listed by the BIOS.\n",
								 boot_cpu_id);
		physid_set(hard_smp_processor_id(), phys_cpu_present_map);
	}

	/*
	 * If we couldn't find a local APIC, then get out of here now!
	 */
1056
	if (!cpu_has_apic) {
Linus Torvalds's avatar
Linus Torvalds committed
1057
1058
1059
		printk(KERN_ERR "BIOS bug, local APIC #%d not detected!...\n",
			boot_cpu_id);
		printk(KERN_ERR "... forcing use of dummy APIC emulation. (tell your hw vendor)\n");
1060
1061
		nr_ioapics = 0;
		return -1;
Linus Torvalds's avatar
Linus Torvalds committed
1062
1063
1064
1065
1066
1067
1068
	}

	/*
	 * If SMP should be disabled, then really disable it!
	 */
	if (!max_cpus) {
		printk(KERN_INFO "SMP mode deactivated, forcing use of dummy APIC emulation.\n");
1069
1070
		nr_ioapics = 0;
		return -1;
Linus Torvalds's avatar
Linus Torvalds committed
1071
1072
	}

1073
1074
	return 0;
}
Linus Torvalds's avatar
Linus Torvalds committed
1075

1076
1077
1078
1079
/*
 * Prepare for SMP bootup.  The MP table or ACPI has been read
 * earlier.  Just do some sanity checking here and enable APIC mode.
 */
1080
void __init smp_prepare_cpus(unsigned int max_cpus)
1081
1082
1083
1084
{
	nmi_watchdog_default();
	current_cpu_data = boot_cpu_data;
	current_thread_info()->cpu = 0;  /* needed? */
1085
	set_cpu_sibling_map(0);
Linus Torvalds's avatar
Linus Torvalds committed
1086

1087
1088
1089
1090
	if (smp_sanity_check(max_cpus) < 0) {
		printk(KERN_INFO "SMP disabled\n");
		disable_smp();
		return;
Linus Torvalds's avatar
Linus Torvalds committed
1091
1092
	}

1093

Linus Torvalds's avatar
Linus Torvalds committed
1094
	/*
1095
	 * Switch from PIC to APIC mode.
Linus Torvalds's avatar
Linus Torvalds committed
1096
	 */
1097
	setup_local_APIC();
Linus Torvalds's avatar
Linus Torvalds committed
1098

1099
1100
1101
1102
	if (GET_APIC_ID(apic_read(APIC_ID)) != boot_cpu_id) {
		panic("Boot APIC ID in local APIC unexpected (%d vs %d)",
		      GET_APIC_ID(apic_read(APIC_ID)), boot_cpu_id);
		/* Or can we switch back to PIC here? */
Linus Torvalds's avatar
Linus Torvalds committed
1103
1104
1105
	}

	/*
1106
	 * Now start the IO-APICs
Linus Torvalds's avatar
Linus Torvalds committed
1107
1108
1109
1110
1111
1112
1113
	 */
	if (!skip_ioapic_setup && nr_ioapics)
		setup_IO_APIC();
	else
		nr_ioapics = 0;

	/*
1114
	 * Set up local APIC timer on boot CPU.
Linus Torvalds's avatar
Linus Torvalds committed
1115
1116
	 */

1117
	setup_boot_APIC_clock();
Linus Torvalds's avatar
Linus Torvalds committed
1118
1119
}

1120
1121
1122
1123
/*
 * Early setup to make printk work.
 */
void __init smp_prepare_boot_cpu(void)
Linus Torvalds's avatar
Linus Torvalds committed
1124
{
1125
1126
1127
	int me = smp_processor_id();
	cpu_set(me, cpu_online_map);
	cpu_set(me, cpu_callout_map);
1128
	per_cpu(cpu_state, me) = CPU_ONLINE;
Linus Torvalds's avatar
Linus Torvalds committed
1129
1130
}

1131
1132
1133
1134
/*
 * Entry point to boot a CPU.
 */
int __cpuinit __cpu_up(unsigned int cpu)
Linus Torvalds's avatar
Linus Torvalds committed
1135
{
1136
1137
	int err;
	int apicid = cpu_present_to_apicid(cpu);
Linus Torvalds's avatar
Linus Torvalds committed
1138

1139
	WARN_ON(irqs_disabled());
Linus Torvalds's avatar
Linus Torvalds committed
1140

1141
1142
1143
1144
1145
1146
1147
1148
	Dprintk("++++++++++++++++++++=_---CPU UP  %u\n", cpu);

	if (apicid == BAD_APICID || apicid == boot_cpu_id ||
	    !physid_isset(apicid, phys_cpu_present_map)) {
		printk("__cpu_up: bad cpu %d\n", cpu);
		return -EINVAL;
	}

1149
1150
1151
1152
1153
1154
1155
1156
	/*
	 * Already booted CPU?
	 */
 	if (cpu_isset(cpu, cpu_callin_map)) {
		Dprintk("do_boot_cpu %d Already started\n", cpu);
 		return -ENOSYS;
	}

1157
	per_cpu(cpu_state, cpu) = CPU_UP_PREPARE;
1158
1159
1160
1161
1162
	/* Boot it! */
	err = do_boot_cpu(cpu, apicid);
	if (err < 0) {
		Dprintk("do_boot_cpu failed %d\n", err);
		return err;
Linus Torvalds's avatar
Linus Torvalds committed
1163
	}
1164

Linus Torvalds's avatar
Linus Torvalds committed
1165
1166
1167
1168
	/* Unleash the CPU! */
	Dprintk("waiting for cpu %d\n", cpu);

	while (!cpu_isset(cpu, cpu_online_map))
1169
		cpu_relax();
1170
1171
1172
	err = 0;

	return err;
Linus Torvalds's avatar
Linus Torvalds committed
1173
1174
}

1175
1176
1177
/*
 * Finish the SMP boot.
 */
1178
void __init smp_cpus_done(unsigned int max_cpus)
Linus Torvalds's avatar
Linus Torvalds committed
1179
{
1180
	smp_cleanup_boot();
Linus Torvalds's avatar
Linus Torvalds committed
1181
	setup_ioapic_dest();
1182
	check_nmi_watchdog();
1183
	time_init_gtod();
1184
}
1185
1186
1187

#ifdef CONFIG_HOTPLUG_CPU

1188
static void remove_siblinginfo(int cpu)
1189
1190
{
	int sibling;
1191
	struct cpuinfo_x86 *c = cpu_data;
1192

1193
1194
1195
1196
1197
1198
1199
1200
1201
	for_each_cpu_mask(sibling, cpu_core_map[cpu]) {
		cpu_clear(cpu, cpu_core_map[sibling]);
		/*
		 * last thread sibling in this cpu core going down
		 */
		if (cpus_weight(cpu_sibling_map[cpu]) == 1)
			c[sibling].booted_cores--;
	}
			
1202
1203
1204
1205
	for_each_cpu_mask(sibling, cpu_sibling_map[cpu])
		cpu_clear(cpu, cpu_sibling_map[sibling]);
	cpus_clear(cpu_sibling_map[cpu]);
	cpus_clear(cpu_core_map[cpu]);
1206
1207
	c[cpu].phys_proc_id = 0;
	c[cpu].cpu_core_id = 0;
1208
	cpu_clear(cpu, cpu_sibling_setup_map);
1209
1210
1211
1212
1213
1214
1215
1216
1217
}

void remove_cpu_from_maps(void)
{
	int cpu = smp_processor_id();

	cpu_clear(cpu, cpu_callout_map);
	cpu_clear(cpu, cpu_callin_map);
	clear_bit(cpu, &cpu_initialized); /* was set by cpu_init() */
1218
	clear_node_cpumask(cpu);
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
}

int __cpu_disable(void)
{
	int cpu = smp_processor_id();

	/*
	 * Perhaps use cpufreq to drop frequency, but that could go
	 * into generic code.
 	 *
	 * We won't take down the boot processor on i386 due to some
	 * interrupts only being able to be serviced by the BSP.
	 * Especially so if we're not using an IOAPIC	-zwane
	 */
	if (cpu == 0)
		return -EBUSY;

1236
1237
	if (nmi_watchdog == NMI_LOCAL_APIC)
		stop_apic_nmi_watchdog(NULL);
1238
	clear_local_APIC();
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251

	/*
	 * HACK:
	 * Allow any queued timer interrupts to get serviced
	 * This is only a temporary solution until we cleanup
	 * fixup_irqs as we do for IA64.
	 */
	local_irq_enable();
	mdelay(1);

	local_irq_disable();
	remove_siblinginfo(cpu);

1252
	spin_lock(&vector_lock);
1253
1254
	/* It's now safe to remove this processor from the online map */
	cpu_clear(cpu, cpu_online_map);
1255
	spin_unlock(&vector_lock);
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
	remove_cpu_from_maps();
	fixup_irqs(cpu_online_map);
	return 0;
}

void __cpu_die(unsigned int cpu)
{
	/* We don't do anything here: idle task is faking death itself. */
	unsigned int i;

	for (i = 0; i < 10; i++) {
		/* They ack this in play_dead by setting CPU_DEAD */
1268
1269
		if (per_cpu(cpu_state, cpu) == CPU_DEAD) {
			printk ("CPU %d is now offline\n", cpu);
1270
1271
			if (1 == num_online_cpus())
				alternatives_smp_switch(0);
1272
			return;
1273
		}
1274
		msleep(100);
1275
1276
1277
1278
	}
 	printk(KERN_ERR "CPU %u didn't die...\n", cpu);
}

1279
static __init int setup_additional_cpus(char *s)
1280
{
1281
	return s && get_option(&s, &additional_cpus) ? 0 : -EINVAL;
1282
}
1283
early_param("additional_cpus", setup_additional_cpus);
1284

1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
#else /* ... !CONFIG_HOTPLUG_CPU */

int __cpu_disable(void)
{
	return -ENOSYS;
}

void __cpu_die(unsigned int cpu)
{
	/* We said "no" in __cpu_disable */
	BUG();
}
#endif /* CONFIG_HOTPLUG_CPU */